GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2006-12 > 1166419054


From: (John Chandler)
Subject: Re: [DNA] Help needed NETWORK phylogenetic tree
Date: Mon, 18 Dec 2006 00:17:34 -0500 (EST)
References: <BAY105-F23F8457BC564A59629CEF7CCCB0@phx.gbl>
In-Reply-To: <BAY105-F23F8457BC564A59629CEF7CCCB0@phx.gbl>(stevenbird1000@hotmail.com)


I wrote:
> >Ok, but bear in mind that the probability distribution is skewed.

Steve wrote:
> This would be due to the smaller than desired sample, no doubt.

No, the skew is inherent in the fact that mutations are rare. I wasn't
saying that the estimate is biased, but rather that the probability
distribution is asymmetric, has a much longer tail on the old side than
on the young side, and has an offset between the median and the mode.

> be strong agreement (within 100 years) between McGee's Utility and NETWOR=
> K
> for the TMRCA. Is that due to a similar rule for handling the TMRCA in b=
> oth
> programs?

Very likely.

> What is the default confidence interval in NETWORK? It just says 700 yea=
> rs;
> is that at 50% CI?

That's supposed to be the standard deviation. However, when the
distribution is skewed, you can't give just one number to convey the
confidence interval. In a gaussian distribution, the interval of
plus-or-minus one standard deviation has about 67% confidence.

> Does the NETWORK program always assume that a
> mutation upward in repeats is later in time?

I think it simply places the node that *has* the mutation (as
determined by the "majority rule") on the end.

John Chandler

This thread: